Photoresist model determination method and device, computer equipment and storage medium

Abstract

The embodiment of the invention discloses a photoresist model determination method and device, computer equipment and a storage medium, and relates to the technical field of photoetching. Comprising the steps that model information of a photoresist model is obtained, the model information comprises basic items of the photoresist model, the basic item number range of the basic items and the parameter value range of parameters in the basic items, and the photoresist model is used for predicting a photoresist pattern formed through photoetching; based on the model information, sampling is carried out to obtain a sample photoresist model, the basic item number of basic items in the sample photoresist model is within the basic item number range, and parameter values of parameters in the basic items are within the parameter value range; and based on a photoresist pattern error corresponding to the sample photoresist model, determining a target basic item number and a target parameter value, the photoresist pattern error being an error between a predicted photoresist pattern predicted by the sample photoresist model and an actual photoresist pattern. According to the scheme provided by the embodiment of the invention, the photoresist pattern prediction accuracy of the photoresist model can be improved.

Description

Technical Field

[0001] This application relates to the field of photolithography technology, and in particular to a method, apparatus, computer equipment, and storage medium for determining a photoresist model. Background Technology

[0002] In the manufacturing of very large-scale integrated circuits, the light from the lithography machine exposes the photoresist on the wafer, and after post-baking and development processes, a photoresist pattern is formed on the wafer.

[0003] In related technologies, in order to obtain a photoresist model that matches the photolithography process, a series of test patterns are designed and exposed on a wafer, and then the experimental measurement data of the exposed patterns on the wafer are used to fit the photoresist model.

[0004] In related technologies, the mathematical form of the photoresist model is usually fixed first, and then a fitting is performed. However, this approach results in a limited search space, causing the photoresist model to perform well only under the test pattern, while performing poorly under other images, thus affecting the lithography quality. Summary of the Invention

[0005] This application provides a method, apparatus, computer device, and storage medium for determining a photoresist model. The technical solution is as follows:

[0006] On one hand, embodiments of this application provide a method for determining a photoresist model, the method comprising:

[0007] Obtain model information of a photoresist model, the model information including the basic terms of the photoresist model, the range of the number of basic terms of the basic terms, and the range of parameter values ​​of the parameters in the basic terms. The photoresist model is used to predict the photoresist pattern formed by photolithography.

[0008] Based on the model information, a sample photoresist model is obtained by sampling. The number of basic items in the basic items of the sample photoresist model is within the range of the number of basic items, and the parameter values ​​of the parameters in the basic items are within the range of the parameter values.

[0009] Based on the photoresist pattern error corresponding to the sample photoresist model, the number of target basic items and the target parameter values ​​are determined. The photoresist pattern error is the error between the predicted photoresist pattern obtained by the sample photoresist model and the actual photoresist pattern.

[0010] On the other hand, embodiments of this application provide a device for determining a photoresist model, the device comprising:

[0011] The acquisition module is used to acquire model information of the photoresist model. The model information includes the basic items of the photoresist model, the range of the number of basic items of the basic items, and the range of parameter values ​​of the parameters in the basic items. The photoresist model is used to predict the photoresist pattern formed by photolithography.

[0012] A sampling module is used to sample a sample photoresist model based on the model information, wherein the number of basic items in the sample photoresist model is within the range of the number of basic items, and the parameter values ​​of the parameters in the basic items are within the range of the parameter values.

[0013] The determination module is used to determine the number of target basic items and the target parameter value based on the photoresist pattern error corresponding to the sample photoresist model. The photoresist pattern error is the error between the predicted photoresist pattern obtained by the sample photoresist model and the actual photoresist pattern.

[0014] On the other hand, embodiments of this application provide a computer device including a processor and a memory, wherein the memory stores at least one computer instruction, which is loaded and executed by the processor to implement the method for determining a photoresist model as described above.

[0015] On the other hand, embodiments of this application provide a computer-readable storage medium storing at least one computer instruction, which is loaded and executed by a processor to implement the method for determining a photoresist model as described above.

[0016] On the other hand, embodiments of this application provide a computer program product, the computer program product including computer instructions stored in a computer-readable storage medium, a processor obtaining the computer instructions from the computer-readable storage medium, and the processor executing the computer instructions to implement the photoresist model determination method as described above.

[0017] When using the photoresist model determination method provided in this application embodiment, it is not necessary to fix the mathematical form of the photoresist model. Instead, the input is model information that includes basic terms, the range of the number of basic terms, and the range of parameter values ​​in the basic terms. Subsequently, the computer device samples a sample photoresist model based on this model information. Based on the error between the predicted photoresist pattern obtained from the sample photoresist model and the actual photoresist pattern, it determines the target number of basic terms from the range of the number of basic terms, that is, determines the mathematical form of the photoresist model, and determines the target parameter value from the range of parameter values, that is, determines the model parameters of the photoresist model. This allows the finally determined photoresist model to take into account both the mathematical form and the model parameters, improving the pattern prediction accuracy of the photoresist model and thus helping to improve the lithography quality. Attached Figure Description

[0018] Figure 1 A schematic diagram of a computer system provided in an exemplary embodiment of this application is shown;

[0019] Figure 2 A flowchart illustrating a method for determining a photoresist model provided in an exemplary embodiment of this application is shown;

[0020] Figure 3 A flowchart illustrating the basic item number optimization process provided in an exemplary embodiment of this application is shown;

[0021] Figure 4 This is a schematic diagram illustrating the model feature clustering process in an exemplary embodiment of this application;

[0022] Figure 5 This is a schematic diagram illustrating an exemplary embodiment of the basic item quantity optimization process of this application;

[0023] Figure 6 A flowchart illustrating the parameter value optimization process provided in an exemplary embodiment of this application is shown.

[0024] Figure 7 A flowchart illustrating a parameter value range scaling process provided in an exemplary embodiment of this application is shown.

[0025] Figure 8 This is a schematic diagram illustrating an exemplary embodiment of the parameter value optimization process of this application;

[0026] Figure 9 This is a comparison chart of photoresist pattern errors in photoresist models under different schemes;

[0027] Figure 10 This is a comparison diagram of the outline of the predicted photoresist pattern and the actual photoresist pattern of the photoresist model provided in the embodiments of this application;

[0028] Figure 11 This is a structural block diagram of a photoresist model determination device provided in another exemplary embodiment of this application;

[0029] Figure 12 A schematic diagram of the structure of a computer device provided in an exemplary embodiment of this application is shown. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0031] For ease of understanding, the terms used in the embodiments of this application will be explained below.

[0032] A photoresist model is a model composed of a series of basic terms related to the photolithography process, used to predict the photoresist pattern formed under specific light intensity distributions. In some embodiments, the basic terms may include at least one of the following: Ag, Mg, Acid, Base, and Grad. The Ag term corresponds to the aerial term (imaging period), the Mg term corresponds to the mask, the Acid and Base terms correspond to the photoacid generator (PAG) and its chemical reaction with the photoresist, and the Grad term corresponds to the gradient process.

[0033] Different base terms have their own base term parameters. In some embodiments, the Ag term includes the standard deviation of the Ag Gaussian distribution; the Mg term includes the standard deviation of the Mg Gaussian distribution; the Acid term includes the standard deviation of the Acid Gaussian distribution and the Acid cutoff threshold; the Base term includes the standard deviation of the Base Gaussian distribution and the Base cutoff threshold; and the Grad term includes the standard deviation of the Grad Gaussian distribution.

[0034] Furthermore, the photoresist model has multiple mathematical forms, which are related to the number of basic terms in each basic term of the photoresist model.

[0035] In an illustrative example, the fundamental terms of the photoresist model can be expressed mathematically as follows:

[0036]

[0037] R = max(RI - t, 0)

[0038] Where R is the photoresist pattern calculated by the photoresist model, and Ag, Mg, Acid, Base, and Grad are the basic terms of the photoresist model; α Ag α Mg α Acid α Baseand α Grad These are the coefficients of the five basic terms in the photoresist model; δ Ag δ Mg δ Acid δ Base δ Grad These are the standard deviations of the Gaussian distribution function G for each term (parameters to be optimized in the model), t, and t', respectively. Acid and t Base These are the threshold values ​​of the cutoff function max for the corresponding terms (parameters to be optimized in the model); I represents the light intensity distribution, i.e., the distribution of the lithography beam on the photoresist in the wafer, which is usually obtained through optical model calculation or experimental measurement. and These are the first-order partial derivatives of the light intensity distribution along the X and Y axes, respectively. conv represents the convolution operation, and M represents the photomask.

[0039] Gaussian Process Regression (GPR) is a probabilistic, non-parametric machine learning method that models the correlations between data points using a Gaussian process. In GPR, data points are assumed to be implementations of a Gaussian process that defines a prior distribution, typically parameterized by the mean and covariance functions (kernel functions). Given a dataset, GPR can provide predictions about unknown data points and estimates of the uncertainty of those predictions.

[0040] Latin Hypercube Sampling (LHS) is a stratified sampling technique that approximates random sampling with fewer iterations, thereby accurately reconstructing the input distribution. The key to LHS is stratifying the input probability distribution and then randomly sampling within each interval or stratum. This ensures that the samples represent the values ​​in each interval, more accurately reflecting the distribution of values ​​in the input probability distribution, resulting in a more uniform sample distribution and more comprehensive sampling.

[0041] Please refer to Figure 1 This illustration shows a schematic diagram of a computer system provided in one embodiment of this application. The computer system may include: a computer device 10 and a lithography device 20.

[0042] Computer device 10 is an electronic device with data computing, processing, and storage functions. This computer device 10 can be either a terminal device or a server. Computer device 10 may include, but is not limited to, electronic devices such as mobile phones, computers, and automated control systems.

[0043] In this embodiment, the computer device 10 runs an Electronic Design Automation (EDA) program or a lithography application. This EDA or lithography application determines the mathematical form of the photoresist model and its parameters based on the model information of the input photoresist model. The determined photoresist model is then used to predict the photoresist pattern formed under the input light intensity distribution.

[0044] The computer device 10 and the lithography device 20 can be connected via a wired or wireless network.

[0045] The photolithography equipment 20 is a device that uses an electron beam to expose photoresist to create patterns. In one possible photolithography process, the photolithography equipment 20 generates an electron beam based on light intensity distribution, controls the movement of the electron beam on the substrate surface based on the target imaging image, and gradually forms the desired target pattern. A photomask refers to a pattern master used in semiconductor manufacturing for selective exposure during photolithography. A substrate refers to a material used to support and carry the target pattern on the photomask. Optionally, the substrate has properties such as high transparency, low coefficient of thermal expansion, and high tensile strength, and may be, but is not limited to, any of the following materials: quartz, glass, and resin, etc. Optionally, the substrate may also include one or more thin films.

[0046] A thin film refers to a functional material covering the surface of a photomask substrate. For example, the substrate surface of a photomask may include one or more of the following thin films: photoresist layer, opaque layer, anti-reflective coating, hard mask layer, etc., and may also be thin films with other functions; this application embodiment does not limit this. The photoresist layer is a photosensitive material covering the substrate of the photomask, used to form a target pattern on the substrate during the photomask manufacturing process. The opaque layer is used to block some light during the photolithography process for manufacturing integrated circuits, so as to form a circuit pattern on the substrate used for manufacturing integrated circuits. The opaque layer can be formed on the substrate of the photomask using chromium, other metals, or metal alloys. The anti-reflective coating is used to reduce or eliminate light reflection on the photomask during the photolithography process for manufacturing integrated circuits, thereby improving the contrast and resolution of the circuit pattern on the substrate used for manufacturing integrated circuits. The anti-reflective coating can be formed on the substrate of the photomask using high refractive index materials such as tantalum oxide and silicon nitride. The hard mask layer is used to protect the accuracy of the target pattern on the substrate of the photomask during subsequent etching or processing. Hard mask layers can be generated on the substrate of the mask using etching-resistant materials such as titanium nitride and tantalum oxide.

[0047] In the process of determining the mathematical form and model parameters of the photoresist model based on the input model information, the computer device 10 uses sampled photoresist models to predict the photoresist pattern. To determine the difference between the predicted photoresist pattern and the actual photoresist pattern produced by lithography, the computer device 10 instructs the lithography equipment 20 to perform lithography and acquires the predicted lithography pattern 11 output by the sample photoresist model, and the actual lithography pattern 21 obtained by the lithography equipment 20. Based on the difference between the predicted lithography pattern 11 and the actual lithography pattern 21, the computer device 10 can determine the mathematical form and model parameters of the photoresist model through iterative sampling. Specifically, for different lithography equipment or lithography processes, the above scheme can be used to determine a photoresist model suitable for that particular lithography equipment or process.

[0048] In the following embodiments, the method for determining the photoresist model is used for Figure 1 The computer device 10 shown is used as an example for explanation.

[0049] Figure 2 A flowchart illustrating a method for determining a photoresist model according to an exemplary embodiment of this application is shown. This embodiment uses this method for... Figure 1 Taking computer device 10 as an example, the method includes the following steps.

[0050] Step 201: Obtain the model information of the photoresist model. The model information includes the basic terms of the photoresist model, the range of the number of basic terms, and the range of parameter values ​​in the basic terms. The photoresist model is used to predict the photoresist pattern formed by photolithography.

[0051] In this embodiment of the application, since it is necessary to take into account both the mathematical form and the coordinated optimization of parameters, the mathematical form of the photoresist model is not fixed in the model information input to the computer device. That is, the number of basic items of each basic item in the photoresist model is not fixed, but a range of basic item numbers is set for the basic items.

[0052] In some embodiments, each basic item corresponds to its own basic item quantity range, and the basic item quantity ranges corresponding to different basic items may be the same or different.

[0053] In an illustrative example, the model information includes basic items Ag, Mg, Acid, Base, and Grad. The number of basic items for each basic item ranges from [1, 4], [2, 5], [1, 3], [1, 5], to [1, 2]. That is, the photoresist model includes a maximum of 4 Ag items and a minimum of 1 Ag item; a maximum of 5 Mg items and a minimum of 2 Mg items; a maximum of 3 Acid items and a minimum of 1 Acid item; a maximum of 5 Base items and a minimum of 2 Base items; and a maximum of 2 Mg items and a minimum of 1 Grad item.

[0054] In addition to specifying the basic items included in the photoresist model and the range of their quantity, the model information also includes the range of parameter values ​​for the parameters in the basic items.

[0055] In an illustrative example, the parameters in the basic terms include δ. Ag δ Mg δ Acid δ Base δ Grad ,t,t Acid and t Base The corresponding parameter ranges are represented as Rδ. Ag ,Rδ Mg ,Rδ Acid ,Rδ Base ,Rδ Grad ,Rt,Rt Acid and Rt Base .

[0056] In some embodiments, the model information may also include the range of basic term coefficients for each basic term, wherein the basic term coefficients corresponding to the basic terms are determined when the number of basic terms and the parameters of the basic terms are determined.

[0057] In an illustrative example, when the basic terms include Ag, Mg, Acid, Base, and Grad terms, the coefficients of the basic terms include α. Ag α Mg α Acid α Base and α Grad .

[0058] Step 202: Based on the model information, sample photoresist model is obtained. The number of basic terms in the sample photoresist model is within the range of basic terms, and the parameter values ​​of the basic terms are within the range of parameter values.

[0059] In one possible implementation, the computer device samples the number of basic items within the range of basic item quantity specified in the model information, and samples the parameter values ​​within the range of parameter values ​​specified in the model information to obtain multiple sample photoresist models, each of which conforms to the specifications of the model information.

[0060] In some embodiments, the sampling process of the sample photoresist model can be executed in multiple rounds, that is, in different iteration rounds, the computer device samples the sample photoresist model based on the sampling requirements of the current iteration round.

[0061] Optionally, the computer device can use random sampling to obtain the sample photoresist model; or, in order to improve the sampling uniformity, the computer device can use Latin hypercube sampling or other sampling methods. The sampling process will be described in detail in the following embodiments.

[0062] Step 203: Based on the photoresist pattern error corresponding to the sample photoresist model, determine the number of target basic items and the target parameter values. The photoresist pattern error is the error between the predicted photoresist pattern obtained by the sample photoresist model and the actual photoresist pattern.

[0063] To evaluate the accuracy of photoresist pattern prediction using a sample photoresist model, in one possible implementation, a computer device uses the sample photoresist model to predict a predicted photoresist pattern, and then performs actual photolithography using a photolithography device to obtain an actual photoresist pattern. The accuracy of the photoresist pattern prediction is then evaluated by comparing the difference between the actual and predicted photoresist patterns. The predicted and actual photoresist patterns correspond to the same lithographic pattern under the same light intensity distribution conditions.

[0064] In some embodiments, the computer device determines the sample photoresist model corresponding to the minimum photoresist pattern error as the target photoresist model obtained through optimization. The number of basic terms of the target photoresist model is the target number of basic terms, and the parameter values ​​of the target photoresist model are the target parameter values.

[0065] It should be noted that the determined target photoresist model also has target basic term coefficients corresponding to each basic term, and these target basic term coefficients are related to the target parameter values ​​and the number of target basic terms.

[0066] In summary, when using the photoresist model determination method provided in this application embodiment, it is not necessary to fix the mathematical form of the photoresist model. Instead, the input is model information that includes basic terms, the range of the number of basic terms, and the range of parameter values ​​in the basic terms. Subsequently, the computer device samples a sample photoresist model based on this model information. Based on the error between the predicted photoresist pattern obtained from the sample photoresist model and the actual photoresist pattern, it determines the target number of basic terms from the range of the number of basic terms, i.e., determines the mathematical form of the photoresist model, and determines the target parameter value from the range of parameter values, i.e., determines the model parameters of the photoresist model. This allows the finally determined photoresist model to take into account both the mathematical form and the model parameters, improving the pattern prediction accuracy of the photoresist model and thus helping to improve the lithography quality.

[0067] Given a defined range of basic terms and parameter values, the search space for optimizing a photoresist model is enormous. To improve the efficiency of the optimization search for the photoresist model, in one possible implementation, the optimization process can be divided into two stages: an optimization stage for the number of basic terms and an optimization stage for the parameter values.

[0068] In the optimization phase for the number of basic items, the computer equipment determines the target number of basic items based on the photoresist pattern error corresponding to the first sample photoresist model. Specifically, the number of basic items in the first sample photoresist model must be within the range of basic item counts, and the parameter values ​​of the basic items must be within the range of parameter values.

[0069] In the parameter optimization phase, after determining the target number of basic items in the basic item quantity optimization phase, the computer equipment fixes the target number of basic items and determines the target parameter values ​​based on the photoresist pattern error corresponding to the second sample photoresist model. Specifically, the number of basic items in the second sample photoresist model is the target number of basic items, and the parameter values ​​of the basic items are within the parameter value range.

[0070] Since the search space for optimizing the number of basic items is smaller than the search space for optimizing the parameter values, optimizing the number of basic items in the first stage can improve the efficiency of optimizing the parameter values ​​in the second stage.

[0071] The following examples illustrate in detail the optimization stages for the number of basic items and the optimization stages for parameter values.

[0072] Figure 3 A flowchart of a basic item quantity optimization process provided by an exemplary embodiment of this application is shown, which may include the following steps.

[0073] Step 301: Based on the range of the number of basic terms, determine multiple model feature spaces, with different model feature spaces having different numbers of basic terms.

[0074] In one possible implementation, the computer device arranges and combines the number of basic items for different basic items based on the range of the number of basic items corresponding to each basic item, thereby obtaining multiple model feature spaces.

[0075] In an illustrative example, if the basic terms of the photoresist model include Ag, Mg, Acid, Base, and Grad terms, and the number of basic terms is in the range of [1, 5], the feature space of the model can be represented as [1, 1, 1, 1, 1], [1, 1, 1, 1, 2], ..., [2, 2, 2, 2, 1], [2, 2, 2, 2, 2], ..., [5, 5, 5, 5, 5]. Here, the model feature space [1, 1, 1, 1, 1] indicates that the number of basic terms Ag, Mg, Acid, Base, and Grad in the photoresist model is all 1. Other model feature spaces are not elaborated upon.

[0076] In some embodiments, the model feature space can be used not only to indicate the number of basic terms but also to indicate the number of parameters in the basic terms.

[0077] In one possible implementation, each row of the model feature space represents a sampled data point, and the number of columns is 5+N. max，Ag +N max，Mg +2N max，Acid +2N max，Base +N max，Grad +1. The first five columns represent the number of basic items for each basic item, N. max，Ag N max，Mg N max，Acid N max，Base and N max，Grad The first five columns represent the maximum number of each basic item, and the last column represents the parameter t. The remaining columns, excluding the first five, represent the parameters included in each basic item, with zeros added to fill any missing parameters in the basic items.

[0078] In an illustrative example, if the maximum number of basic terms for each basic term is 2, the model feature space can be represented as:

[0079] [1, 1, 1, 1, 1, δ Ag1 ,0,δ Mg1 ,0,δ Acid1 ,0,t Acid1 ,0,δ Base1 ,0,t Base1 ,0,δ Grad1 [0, t]

[0080] ...

[0082] Of course, the model feature space can also take other forms than those described above, as long as the model feature space can represent the number of basic terms for each basic term. This embodiment does not limit this.

[0083] Step 302: Based on the parameter value range, the first sample photoresist model is obtained by sampling in each model feature space. The first sample photoresist model is obtained by random sampling or Latin hypercube sampling.

[0084] In one possible implementation, for each model feature space, the computer device samples the parameter values ​​of the parameters in the basic terms under the model feature space based on the parameter value range in the basic terms, thereby obtaining multiple first sample photoresist models under the model feature space. Each first sample photoresist model has a given number of basic terms and parameter values.

[0085] Optionally, the computer device can use random sampling to sample the first sample photoresist model in the model feature space.

[0086] Optionally, in order to achieve more uniform and comprehensive sampling in the model feature space, computer devices can perform sampling in the model feature space using the Latin hypercube sampling method.

[0087] In some embodiments, under the Latin hypercube sampling method, the computer device divides each dimension (the dimension corresponding to the parameters in the basic terms) in the model feature space into L non-overlapping intervals of equal length, and randomly samples a point in each interval of each dimension. Combining the randomly sampled points in each dimension yields the first sample photoresist model.

[0088] Based on the examples in the steps above, for the model feature space [1, 1, 1, 1, 1, δ] Ag1 ,0,δ Mg1 ,0,δ Acid1 ,0,t Acid1 ,0,δ Base1 ,0,t Base1 ,0,δ Grad1 [0, t], computer equipment will δ Ag1 δ Mg1 δ Acid1 t Acid1 δ Base1 t Base1 δ Grad1 The parameter values ​​of t are divided into L intervals of equal length. Samples are taken in each interval, and the sampled points are combined to obtain a first sample photoresist model with a number of 1 for each of the basic terms.

[0089] Of course, in addition to the two sampling methods mentioned above, computer devices may also use other sampling methods, and this application embodiment does not limit this.

[0090] Step 303: Divide the first sample photoresist model into a first model set and a second model set.

[0091] In one possible implementation, since the cost of photolithography using photolithography equipment is high, in order to reduce the actual number of photolithography steps in the photoresist model determination process and thus reduce the cost of determining the photoresist model, in this embodiment of the application, the computer device uses an active learning method to determine the photoresist model.

[0092] In the process of determining the photoresist model using active learning, the computer learns the mapping relationship between the mathematical form of the photoresist model parameters and the prediction error based on samples in a small sample set. This learned mapping relationship is then applied to samples in a large sample set. High-quality samples are selected from the large sample set and added to the small sample set. The learning process is repeated based on the updated samples in the small sample set. Through multiple iterations, the learned mapping relationship is optimized, ultimately yielding the optimal samples.

[0093] In one possible implementation, after the first sample photoresist model is sampled, the computer device first divides the sampled first sample photoresist model into a first model set and a second model set.

[0094] Optionally, the number of first sample photoresist models in the first model set is less than the number of first sample photoresist models in the second model set.

[0095] In some embodiments, the initial samples in the first model set are randomly selected from a large sample set (i.e., the sampled first photoresist model) or a small subset of samples selected based on a heuristic method. Furthermore, during the iterative learning process, high-quality samples from the second model set are added to the first model set.

[0096] To improve the quality of the initial samples in the first model set, optionally, the computer device may use clustering to extract representative first sample photoresist models as the initial samples in the first model set. This process may include the following steps:

[0097] Step 303A: Cluster the model features of the first sample photoresist model to obtain multiple clusters.

[0098] The model features of the first sample photoresist model can be represented by the number of basic terms and parameter values. Optionally, these model features can be in the form of a model feature vector for subsequent clustering. The different dimensions of the model feature vector correspond to the number of basic terms and the parameter values ​​of the parameters in each basic term of the first sample photoresist model.

[0099] In one possible implementation, the computer device performs clustering using the K-means algorithm, given a number of cluster centers N. The core idea of ​​the K-means clustering algorithm is to divide the data into N clusters based on the distance between data points (e.g., Euclidean distance), minimizing the distance between data points within a cluster and maximizing the distance between clusters. During the clustering process, the squared error (error) is minimized by continuously changing the positions of the cluster centers. This process can be represented as:

[0100]

[0101] Where N represents the number of cluster centers, x is the model feature vector, and μ i Represents the location of the i-th cluster center, |C i | represents the number of data samples in the i-th cluster.

[0102] When the distance between the N cluster centers no longer changes, that is, when the squared error converges, the K-means algorithm outputs the locations of the N cluster centers and the cluster to which each sample belongs.

[0103] Step 303B: Add the first sample photoresist model corresponding to the model feature closest to the cluster center in each cluster to the first model set.

[0104] For each cluster obtained by clustering, the computer device determines the model feature that is closest to the cluster center in the cluster, and then adds the first sample photoresist model corresponding to the model feature to the first model set.

[0105] For example, if N clusters are obtained through clustering, the computer device determines N model features from the N clusters and adds the N first sample photoresist models corresponding to the N model features to the first model set.

[0106] like Figure 4 As shown, the computer device clusters the model features into multiple clusters 41, and adds the first sample photoresist model corresponding to the model feature closest to the cluster center in each cluster 41 to the first model set.

[0107] Step 303C: Add the first sample photoresist model, which is not in the first model set, to the second model set.

[0108] Furthermore, the computer device adds the first sample photoresist model, which is not in the first model set, to the second model set as the initial sample in the second model set.

[0109] Of course, in addition to using the above clustering method to determine the initial samples in the first model set, the computer device can also use other methods (the quality of the determined initial samples may vary). For example, for each combination of basic items, the first sample photoresist model representing the combination of basic items can be extracted based on the median of the parameter values. This application embodiment does not constitute a limitation in this regard.

[0110] Step 304: Based on the first sample photoresist model in the first model set, determine the coefficients of each basic term using the least squares method.

[0111] Since the sampling process only sampled the number of basic items and the parameter values ​​of the parameters in the basic items, while the photoresist model also includes the coefficients of each basic item, the computer equipment still needs to further determine the coefficients of each basic item based on the first sample photoresist model in the first model set.

[0112] As can be seen from the mathematical expression of the above photoresist model, when the photoresist pattern (i.e., R) and the light intensity distribution (i.e., I) are consistent, the coefficients of the basic terms are linearly related to the values ​​of each basic term (determined based on the parameters in the basic terms and the light intensity distribution). Therefore, the computer device can use the least squares method to determine the coefficients of each basic term.

[0113] In some embodiments, the computer device uses a constrained least squares method to determine the coefficients of the basic terms. The constraint condition is used to constrain the range of each basic term coefficient, ensuring that the determined basic term coefficients fall within the constrained range.

[0114] In some embodiments, when the basic terms include Ag, Mg, Acid, Base, and Grad, the computer device determines the basic term coefficients α using the least squares method based on the coefficient ranges corresponding to Ag, Mg, Acid, Base, and Grad. Ag α Mg α Acid α Base and α Grad .

[0115] Step 305: The predicted photoresist pattern is obtained by using the first sample photoresist model with the basic term coefficients.

[0116] The computer device applies the basic term coefficients determined in step 304 above to each of the first sample photoresist models in the first model set. Since the number of basic terms, the basic term coefficients, and the parameters in the basic terms of the first sample photoresist models have all been determined, the computer device can predict the photoresist pattern based on the given light intensity distribution using the first sample photoresist models, and obtain the predicted photoresist pattern.

[0117] The process of predicting the photoresist pattern is a mathematical calculation process, which requires actual photolithography using photolithography equipment.

[0118] Step 306: The error between the predicted photoresist pattern and the actual photoresist pattern is determined as the photoresist pattern error corresponding to the first sample photoresist model.

[0119] To evaluate the prediction accuracy of the first sample photoresist model, the computer equipment needs to perform actual photolithography using a photolithography device to obtain an actual photoresist pattern. Specifically, the light intensity distribution and photolithography pattern used for prediction based on the first sample photoresist model must be consistent with the light intensity distribution and photolithography pattern used in the actual photolithography process.

[0120] After obtaining the actual photoresist pattern, the computer device calculates the error between the predicted photoresist pattern and the actual photoresist pattern for each first sample photoresist in the first model set.

[0121] In one possible implementation, the computer device calculates the edge distance error (EDE) between the predicted photoresist pattern and the actual photoresist pattern. This error can be expressed as:

[0122]

[0123] Where L is the perimeter of the lithographic pattern, Wafer model and Wafer true These are the predicted photoresist pattern obtained from the first sample photoresist model and the actual photoresist pattern obtained from actual photolithography by the photolithography equipment.

[0124] Of course, in addition to using edge distance error, computer devices may also use other error algorithms to determine the edge differences between patterns, and this application embodiment does not limit this.

[0125] Step 307: Based on the model features of the first sample photoresist model in the first model set and the photoresist pattern error corresponding to the first sample photoresist model in the first model set, construct the mapping relationship between model features and errors. The model features include the number of basic terms and parameter values.

[0126] Since the prediction accuracy of the photoresist model is related to the corresponding photoresist pattern error, and it is necessary to select high-quality first sample photoresist models from the second model set based on the model features of the first sample photoresist models in the second model set, the computer device constructs a mapping relationship between model features and errors based on the model features of the first sample photoresist models in the first model set and the corresponding photoresist pattern errors of each first sample photoresist model.

[0127] In one possible implementation, the computer device uses a Gaussian process regression model to construct the above mapping relationship. Here, a Gaussian process is a set of random variables that satisfy a joint Gaussian distribution. The overall root mean square error of the dataset can be described by a Gaussian process model, in the form: Error(x) ~ GP(m(x), k(x, x′)). Here, x and x′ are the model feature vectors of two different photoresist models, and m(x) and k(x, x′) are their mean function and kernel function (used to evaluate similarity), respectively.

[0128] Gaussian process regression is a nonparametric regression algorithm designed to establish a function distribution consistent with the training set. It uses an n-dimensional squared exponential function as the kernel function, expressed as follows: Here, x and x′ are the model feature vectors of two different photoresist models, x i It is the i-th parameter in the model feature vector x, x′ i is the i-th parameter in the model feature vector x′, where n is the dimension of x. These are hyperparameters obtained through maximum likelihood estimation.

[0129] In one possible implementation, the computer device constructs a Gaussian process regression model based on the model features of a first sample photoresist model in a first model set, and the corresponding photoresist pattern error. The input to the Gaussian process regression model is the model features, and the output is the distribution of the photoresist pattern error, which is represented by the mean (μ(x)) and uncertainty (σ(x)). That is, the Gaussian process regression model does not directly output the photoresist error based on the input model features, but rather outputs the mean and uncertainty of the photoresist model error.

[0130] Optionally, the distribution of photoresist pattern error is represented as GP(μ(x), σ(x)).

[0131] Step 308: Based on the mapping relationship, select the first sample photoresist model from the second model set and update it to the first model set.

[0132] Furthermore, the computer device will learn the mapping relationship between model features and errors from the first model set, perform model optimization in the second model set, and then add the optimized first sample photoresist model to the first model set.

[0133] Optionally, the computer device determines the photoresist pattern error corresponding to each first sample photoresist model in the second model set based on the mapping relationship, thereby selecting the first sample photoresist model based on the photoresist pattern error.

[0134] In one possible implementation, when the mapping relationship is constructed using a Gaussian process regression model, the computer device uses the Gaussian process regression model for optimization.

[0135] Optionally, optimization using a Gaussian process regression model may include the following steps:

[0136] Step 308A: Based on the Gaussian process regression model, determine the mean and uncertainty corresponding to the model characteristics of the first sample photoresist model in the second model set.

[0137] Optionally, the computer device inputs the model features corresponding to the first sample photoresist model in the second model set into the Gaussian process regression model to obtain the mean μ(x) and uncertainty σ(x) corresponding to the model features output by the Gaussian process regression model.

[0138] Step 308B: Based on the mean and uncertainty, determine the expected improvement value corresponding to the first sample photoresist model in the second model set. The expected improvement value is used to characterize the quality of the first sample photoresist model.

[0139] In one possible implementation, the computer device further determines the acquisition function of each model feature based on the mean and uncertainty, and optimizes based on the acquisition function.

[0140] In some embodiments, to better balance local and global searches, the obtained function is characterized by expected improvement (EI). The expression for expected improvement is:

[0141]

[0142] Where z = (f min -μ(x)) / σ(x), f min It is the minimum error in the first model set, Φ(z) and These are the cumulative density function and probability density function of the normal distribution function, respectively.

[0143] The first term (σ(x)zΦ(z)) tends to find a result better than the current minimum error, representing a local search; the second term... This involves finding regions with greater uncertainty in the remaining parameter space, representing a global search. Therefore, using expectation boosting as the acquisition function can better balance local and global searches.

[0144] In some embodiments, the computer device determines the expected improvement value corresponding to the first sample photoresist model based on the mean and uncertainty, wherein a larger expected improvement value indicates a higher quality of the first sample photoresist model, and a smaller expected improvement value indicates a lower quality of the first photoresist model.

[0145] Step 308C: Based on the expected improvement value, select the first sample photoresist model from the second model set and add it to the first model set.

[0146] In one possible implementation, the computer device selects the first sample photoresist models with the highest expected improvement values ​​(M) based on the descending order of the expected improvement values ​​corresponding to each first sample photoresist model in the second model set, and adds them to the first model set.

[0147] Step 309: Update the mapping relationship based on the iteratively updated first model set.

[0148] Steps 304 to 308 above constitute one round of iterative update process. Since the newly added first sample photoresist model in the first model set will affect the mapping relationship between model features and errors, the computer device needs to update the mapping relationship between model features and errors based on the first sample photoresist model in the first model set after iterative update.

[0149] The process of updating the mapping relationship involves repeating steps 304 to 307, which means re-determining the coefficients of the basic terms using the least squares method, recalculating the photoresist pattern error between the predicted and actual photoresist patterns, and thus reconstructing the mapping relationship based on the model features and the photoresist pattern error. This embodiment will not elaborate further.

[0150] After completing the mapping update, the computer device needs to optimize again in the second model set based on the updated mapping, and add the optimized first sample photoresist model to the first model set (the process is similar to step 308, and will not be described in detail in this implementation), thereby triggering the mapping update again. That is, the computer device needs to iteratively execute steps 304 to 309 above, continuously optimizing and updating the first model set.

[0151] Step 310: When the number of iterations of the first model set reaches the first iteration threshold, the number of basic terms of the first sample photoresist model corresponding to the minimum photoresist pattern error in the first model set is determined as the target number of basic terms.

[0152] In one possible implementation, after each iteration update of the first model set, the computer device increments the iteration update count by one, and stops the iteration update when the iteration update count reaches a first iteration count threshold.

[0153] Optionally, the threshold for the first iteration count can be set when inputting model information. For example, the threshold for the first iteration count can be set to 5 times, or 10 times, etc.

[0154] After stopping iterative updates, the computer equipment determines the first sample photoresist model with the minimum photoresist pattern error, and determines the number of basic terms of the first sample photoresist model as the target number of basic terms.

[0155] Through steps 301 to 310 above, the computer equipment completes the optimization of the number of basic items, that is, determines the mathematical form of the photoresist model.

[0156] In an illustrative example, such as Figure 5 As shown, the computer device first determines the initial model set through cluster sampling, that is, based on the clustering results of the model features (number of basic terms and parameter values) of the first sample photoresist model 501, dividing the first sample photoresist model 501 into a first model set 502 and a second model set 503. After the model set initialization is completed, the computer device determines the basic term coefficients based on the first sample photoresist model 501 in the first model set 502 using the constrained least squares method. Then, it uses the first sample photoresist model 501 with the basic term coefficients to predict the photolithography pattern, obtains the predicted photoresist pattern, and performs actual photolithography using photolithography equipment to obtain the actual photoresist pattern, thereby determining the photoresist pattern error.

[0157] If the number of iterations of the first model set 502 has not reached the first iteration threshold, the computer device trains a Gaussian process regression model 504 based on the model features of the models in the current first model set 502 and the photoresist pattern error, and applies the Gaussian process regression model 504 to the model optimization process.

[0158] During the model optimization process, the computer applies the Gaussian process regression model 504 to the models in the second model set 503 to obtain the mean and uncertainty representing the error distribution 505. Further, based on the mean and uncertainty, the computer determines the expected improvement value 506 corresponding to the first sample photoresist model 501 in the second model set 503, thereby maximizing the expected improvement value 506 (i.e., argmax). x Multiple first-sample photoresist models 501 (i.e., argmax) of EI(x) xEI(x)) is added to the first model set 502.

[0159] When the number of iterations of the first model set 502 reaches the first iteration threshold, the computer device stops iterating and updating the first model set 502, and determines the number of basic terms corresponding to the minimum photoresist pattern error in the current first model set 502 as the target number of basic terms.

[0160] In this embodiment, the computer device uses cluster sampling to initialize the first model set and the second model set, so that the initial first sample photoresist model in the first model set can reflect the model characteristics of a cluster, which helps to improve the initialization quality of the first model set, and thus improves the quality of subsequent iterative updates of the first model set.

[0161] In addition, the computer equipment uses a Gaussian process regression model to construct a mapping relationship between model features and photoresist pattern errors, thereby using this mapping relationship to predict the error distribution of sample photoresist models in the second model set, which helps to reduce the amount of computation in the model optimization process.

[0162] Furthermore, when calculating the expected improvement value based on the mean and uncertainty of the characterization error distribution, the computer equipment takes into account both local and global searches, which improves the quality of the sample photoresist model obtained by optimization in the second model set, and thus improves the accuracy of the target number of basic terms determined in the optimization stage of the number of basic terms.

[0163] Figure 6 A flowchart of a parameter value optimization process provided by an exemplary embodiment of this application is shown, which may include the following steps.

[0164] Step 601: Determine the model feature space based on the number of target basic terms.

[0165] Unlike the model feature space determined in the basic term quantity optimization stage, which contains combinations of different basic term quantities, in the parameter value optimization stage, the computer device determines the model feature space based on the target basic term quantity obtained in the basic term quantity optimization stage.

[0166] In an illustrative example, the fundamental terms of the photoresist model include Ag, Mg, Acid, Base, and Grad terms, and the number of the identified target fundamental terms is 1, 2, 1, 1, and 1, respectively. The feature space of the model determined by the computer device can then be represented as:

[0167] [1, 2, 1, 1, 1, δ] Ag1 δ Mg1 δ Mg2 δ Acid1 , t Acid1 δBase1 , t Base1 δ Grad1 ,t]

[0168] Step 602: Based on the parameter value range, a second sample photoresist model is obtained by sampling in the model feature space. The second sample photoresist model is obtained by random sampling or Latin hypercube sampling.

[0169] In one possible implementation, the computer device samples the parameter values ​​of the parameters in the basic terms under the model feature space based on the parameter value range of the basic terms, thereby obtaining multiple second sample photoresist models under the model feature space. The number of basic terms in the basic terms of each second sample photoresist model is the same as the target number of basic terms, and the parameter values ​​of different second sample photoresist models are different.

[0170] Optionally, the computer device can use random sampling to sample the second sample photoresist model in the model feature space.

[0171] Optionally, to achieve more uniform and comprehensive sampling within the model feature space, the computer device can perform sampling in the model feature space using the Latin hypercube sampling method. The specific process of sampling the second sample photoresist model can refer to the sampling process of the first sample photoresist model, and will not be elaborated upon in this embodiment.

[0172] Step 603: Divide the second sample photoresist model into a third model set and a fourth model set.

[0173] Similar to the optimization process for the number of basic items, the computer equipment uses an active learning approach to optimize the parameters of the photoresist model. Accordingly, the computer equipment first divides the sampled second-sample photoresist model into a third model set and a fourth model set.

[0174] Optionally, the number of second-sample photoresist models in the third model set is less than the number of second-sample photoresist models in the fourth model set.

[0175] In some embodiments, the initial samples in the third model set are randomly selected from a large sample set (i.e., the sampled second sample photoresist models) or a small subset of samples selected based on a heuristic method. Furthermore, during the iterative learning process, high-quality samples from the fourth model set are added to the third model set.

[0176] To improve the quality of the initial samples in the third model set, optionally, the computer device may use clustering to extract representative second-sample photoresist models as the initial samples in the third model set. This process may include the following steps:

[0177] Step 603A: Cluster the model features of the second sample photoresist model to obtain multiple clusters.

[0178] In some embodiments, the model features include the number of basic terms and the parameter values ​​of the parameters in the basic terms.

[0179] In other embodiments, since the number of basic terms is consistent across all second-sample photoresist models, the model features of the second-sample photoresist models can be represented by parameter values. Optionally, these model features can be in the form of model feature vectors for subsequent clustering. The different dimensions of the model feature vectors correspond to the parameter values ​​of the parameters in each basic term of the second-sample photoresist model.

[0180] Optionally, the computer device can use the K-means algorithm to cluster the model features to obtain multiple clusters. The feature clustering process can refer to step 303A above, and will not be described in detail here.

[0181] Step 603B: Add the second sample photoresist model corresponding to the model feature closest to the cluster center in each cluster to the third model set.

[0182] For each cluster obtained by clustering, the computer device determines the model feature that is closest to the cluster center in the cluster, and then adds the second sample photoresist model corresponding to the model feature to the third model set.

[0183] For example, if N clusters are obtained through clustering, the computer device determines N model features from the N clusters and adds the N second sample photoresist models corresponding to the N model features to the third model set.

[0184] Step 603C: Add the second sample photoresist model, which is not in the third model set, to the fourth model set.

[0185] Furthermore, the computer device adds the second sample photoresist model, which is not in the third model set, to the fourth model set as the initial sample in the fourth model set.

[0186] Step 604: Based on the second sample photoresist model in the third model set, determine the coefficients of each basic term using the least squares method.

[0187] Similar to the process of determining the coefficients of basic terms in the optimization stage of the number of basic terms, given the number of basic terms and parameter values, the computer equipment uses the least squares method to determine the coefficients of each basic term.

[0188] In some embodiments, the computer device uses a constrained least squares method to determine the coefficients of the basic terms. The constraint condition is used to constrain the range of each basic term coefficient, ensuring that the determined basic term coefficients fall within the constrained range.

[0189] Optionally, the computer equipment determines the coefficients of the basic terms based on the same constraints during the optimization of the number of basic terms and the optimization of parameter values.

[0190] Step 605: The predicted photoresist pattern is obtained by using the second sample photoresist model with the basic term coefficients.

[0191] The computer device applies the basic term coefficients determined in step 604 above to each of the second sample photoresist models in the third model set. Since the number of basic terms, the basic term coefficients, and the parameters in the basic terms of the second sample photoresist models have all been determined, the computer device can predict the photoresist pattern based on the given light intensity distribution using the second sample photoresist models, and obtain the predicted photoresist pattern.

[0192] The process of predicting the photoresist pattern is a mathematical calculation process, which requires actual photolithography using photolithography equipment.

[0193] Step 606: The error between the predicted photoresist pattern and the actual photoresist pattern is determined as the photoresist pattern error corresponding to the second sample photoresist model.

[0194] To evaluate the prediction accuracy of the second sample photoresist model, the computer equipment needs to perform actual photolithography using a photolithography device to obtain an actual photoresist pattern. Specifically, the light intensity distribution and photolithography pattern used for prediction based on the second sample photoresist model must be consistent with the light intensity distribution and photolithography pattern used in the actual photolithography process.

[0195] After obtaining the actual photoresist pattern, the computer device calculates the error between the predicted photoresist pattern and the actual photoresist pattern for each second sample photoresist in the third model set.

[0196] In one possible implementation, the computer device calculates the edge distance error between the predicted photoresist pattern and the actual photoresist pattern, or other error algorithms used to determine edge differences between images; this embodiment does not limit this to a specific method.

[0197] Step 607: Based on the model features of the second sample photoresist model in the third model set and the photoresist pattern error corresponding to the second sample photoresist model in the third model set, construct the mapping relationship between features and errors. The model features include the number of basic terms and parameter values.

[0198] Since the prediction accuracy of the photoresist model is related to the corresponding photoresist pattern error, and it is necessary to select high-quality second sample photoresist models from the fourth model set based on the model features of the second sample photoresist models in the third model set, the computer device constructs a mapping relationship between model features and errors based on the model features of the second sample photoresist models in the third model set and the corresponding photoresist pattern errors of the second sample photoresist models.

[0199] In one possible implementation, the computer device uses a Gaussian process regression model to construct the above mapping relationship.

[0200] Optionally, the computer device constructs a Gaussian process regression model based on the model features of the second sample photoresist model in the third model set, and the corresponding photoresist pattern error. The input of the Gaussian process regression model is the model features, and the output is the distribution of the photoresist pattern error, which is represented by the mean (μ(x)) and uncertainty (σ(x)). That is, the Gaussian process regression model does not directly output the photoresist error based on the input model features, but rather outputs the mean and uncertainty of the photoresist model error.

[0201] Optionally, the distribution of photoresist pattern error is represented as GP(μ(x), σ(x)).

[0202] Step 608: Based on the mapping relationship, select the second sample photoresist model from the fourth model set and update it to the third model set.

[0203] Furthermore, the computer device will learn the mapping relationship between model features and errors from the third model set, perform model optimization in the fourth model set, and then add the optimized second sample photoresist model to the third model set.

[0204] Optionally, the computer device determines the photoresist pattern error corresponding to each second sample photoresist model in the fourth model set based on the mapping relationship, and then selects the second sample photoresist model based on the photoresist pattern error.

[0205] In one possible implementation, when the mapping relationship is constructed using a Gaussian process regression model, the computer device uses the Gaussian process regression model for optimization.

[0206] Optionally, optimization using a Gaussian process regression model may include the following steps:

[0207] Step 608A: Based on the Gaussian process regression model, determine the mean and uncertainty corresponding to the model characteristics of the second sample photoresist model in the fourth model set.

[0208] Optionally, the computer device inputs the model features corresponding to the second sample photoresist model in the fourth model set into the Gaussian process regression model to obtain the mean μ(x) and uncertainty σ(x) corresponding to the model features output by the Gaussian process regression model.

[0209] Step 608B: Based on the mean and uncertainty, determine the expected improvement value corresponding to the second sample photoresist model in the fourth model set. The expected improvement value is used to characterize the quality of the second sample photoresist model.

[0210] In one possible implementation, the computer device further determines the acquisition function of each model feature based on the mean and uncertainty, and optimizes based on the acquisition function.

[0211] In some embodiments, to better balance local and global searches, the obtained function is characterized by expected improvement (EI). The expression for expected improvement is:

[0212]

[0213] Where z = (f min -μ(x)) / σ(x), f min It is the minimum error in the third model set, Φ(z) and These are the cumulative density function and probability density function of the normal distribution function, respectively.

[0214] The first term (σ(x)zΦ(z)) tends to find a result better than the current minimum error, representing a local search; the second term... This involves finding regions with greater uncertainty in the remaining parameter space, representing a global search. Therefore, using expectation boosting as the acquisition function can better balance local and global searches.

[0215] In some embodiments, the computer device determines the expected improvement value corresponding to the second sample photoresist model based on the mean and uncertainty, wherein a larger expected improvement value indicates a higher quality of the second sample photoresist model, and a smaller expected improvement value indicates a lower quality of the second photoresist model.

[0216] Step 608C: Based on the expected improvement value, select the second sample photoresist model from the fourth model set and add it to the third model set.

[0217] In one possible implementation, the computer device selects the top M second sample photoresist models with the highest expected improvement values ​​based on the descending order of the expected improvement values ​​corresponding to each second sample photoresist model in the fourth model set, and adds them to the third model set.

[0218] Step 609: Update the mapping relationship based on the updated third model set.

[0219] Steps 604 to 608 above constitute one round of iterative update process. Since the newly added second sample photoresist model in the third model set will affect the mapping relationship between model features and errors, the computer device needs to update the mapping relationship between model features and errors based on the second sample photoresist model in the iteratively updated third model set.

[0220] The process of updating the mapping relationship involves repeating steps 604 to 607, which means re-determining the coefficients of the basic terms using the least squares method, recalculating the photoresist pattern error between the predicted and actual photoresist patterns, and thus reconstructing the mapping relationship based on the model features and the photoresist pattern error. This embodiment will not be elaborated upon further.

[0221] After completing the mapping update, the computer device needs to optimize again in the fourth model set based on the updated mapping, and add the optimized second sample photoresist model to the third model set (the process is similar to step 608, and will not be repeated in this implementation), thereby triggering the mapping update again. That is, the computer device needs to iteratively execute steps 604 to 609 above, continuously optimizing and updating the third model set.

[0222] Step 610: If the iteration termination condition is met, determine the target parameter value by the parameter value of the second sample photoresist model corresponding to the minimum photoresist pattern error in the third model set.

[0223] Optionally, the iteration termination condition may include at least one of an iteration count condition and an error condition. The iteration count condition refers to stopping the iteration update when the number of iterations in the third model set reaches a threshold; the error condition refers to stopping the iteration update when the minimum photoresist pattern error in the third model set is less than an error threshold.

[0224] After stopping the iterative update, the computer equipment determines the second sample photoresist model with the minimum photoresist pattern error, and determines the parameter values ​​of the basic terms in the second sample photoresist model as the target parameter values, and determines the basic term coefficients corresponding to the basic terms as the target basic term coefficients.

[0225] Through steps 601 to 610 above, the computer device completes the parameter value optimization, that is, determines the parameter values ​​(and basic coefficients) of the photoresist model.

[0226] In this embodiment, the computer device uses cluster sampling to initialize the third model set and the fourth model set, so that the initial second sample photoresist model in the third model set can reflect the model characteristics of a cluster, which helps to improve the initialization quality of the third model set, and thus improves the quality of subsequent iterative updates of the third model set.

[0227] In addition, the computer equipment uses a Gaussian process regression model to construct a mapping relationship between model features and photoresist pattern errors, thereby using this mapping relationship to predict the error distribution of sample photoresist models in the fourth model set, which helps to reduce the amount of computation in the model parameter optimization process.

[0228] Furthermore, when calculating the expected improvement value based on the mean and uncertainty of the characterization error distribution, the computer equipment takes into account both local and global searches, which improves the quality of the sample photoresist model obtained by optimization in the fourth model set, and thus improves the accuracy of the target parameter values ​​determined in the parameter value optimization stage.

[0229] Compared to the number of basic terms, since there are more parameters in the basic terms and a larger search space for parameter values, to avoid inefficiency due to an excessively large search range or getting trapped in a local optimum due to an excessively small search range during parameter value optimization, in one possible implementation, during the sampling of the second sample photoresist model in the model feature space, the computer device samples the second sample photoresist model in the model feature space based on the parameter value range and through range scaling. For example... Figure 7 As shown, obtaining the second sample photoresist model by sampling in the model feature space through range scaling can include the following steps:

[0230] Step 701: Based on the parameter value sub-range, the second sample photoresist model is obtained by sampling in the model feature space, and the parameter value sub-range belongs to the parameter value range.

[0231] Here, a parameter value sub-range is a subset of the parameter value range in the model information. Specifically, the maximum value of a parameter value sub-range is less than or equal to the maximum value of the parameter value range, and the minimum value of a parameter value sub-range is greater than or equal to the minimum value of the parameter value range. For example, if the parameter value range of a basic item is 0-100, then the maximum value of this parameter value sub-range is less than or equal to 100, and the minimum value is greater than or equal to 0. For instance, this parameter value sub-range might be 0-50.

[0232] Step 702: When the number of iterations of the third model set reaches the threshold of the second iteration, and the minimum photoresist pattern error in the third model set remains unchanged, the sub-range of parameter values ​​is narrowed, and the narrowed sub-range of parameter values ​​belongs to the parameter value range.

[0233] In one possible implementation, after each iteration of updating the third model set, the computer device determines the minimum photoresist pattern error in the updated third model set. If the minimum photoresist pattern error of the third model set remains unchanged after the number of iterations reaches a second iteration threshold, it indicates that multiple iterations have not found better basic parameters. At this point, the computer device needs to narrow down the sub-range of parameter values ​​(equivalent to narrowing the search grid) and resample the second sample photoresist model within the narrowed sub-range of parameter values.

[0234] Optionally, "minimum photoresist pattern error remains unchanged" can mean that the second sample photoresist model with the minimum photoresist pattern error remains unchanged.

[0235] Optionally, the computer device can reduce the range of parameter values ​​to one-nth of the original range. For example, n is 2.

[0236] It should be noted that during the process of narrowing down the range of parameter values, it is necessary to ensure that the narrowed range of parameter values ​​still falls within the parameter value range.

[0237] In one possible method of narrowing down the subrange of parameter values, the computer device narrows down the subrange of parameter values ​​by centering on the parameter value of the second sample photoresist model corresponding to the smallest photoresist pattern error in the third model set.

[0238] For illustrative purposes, if the parameter value range is 0-40, and the parameter value of the second sample photoresist model corresponding to the minimum photoresist pattern error is 15, the computer equipment determines that the reduced parameter value range is 5-25.

[0239] Step 703: When the number of iterations of the third model set reaches the threshold of the second iteration, and the minimum photoresist pattern error in the third model set decreases, the sub-range of parameter values ​​is enlarged, and the enlarged sub-range of parameter values ​​belongs to the parameter value range.

[0240] In one possible implementation, if the minimum photoresist pattern error of the third model set decreases when the number of iterations reaches the second iteration threshold, it indicates that the optimal solution has been found in the current search space after multiple iterations. At this point, the computer device needs to enlarge the subrange of parameter values ​​(equivalent to expanding the search grid) and resample the second sample photoresist model within the enlarged subrange of parameter values.

[0241] Optionally, the computer equipment can magnify the range of parameter values ​​by a factor of n. For example, n is 2.

[0242] It should be noted that during the process of expanding the range of parameter values, it is necessary to ensure that the expanded range of parameter values ​​still falls within the parameter value range.

[0243] In one possible method of amplifying the subrange of parameter values, the computer device amplifies the subrange of parameter values ​​centered on the parameter value of the second sample photoresist model corresponding to the smallest photoresist pattern error in the third model set.

[0244] For illustration, if the parameter value range is 0-40, and the parameter value of the second sample photoresist model corresponding to the minimum photoresist pattern error is 20, the computer device determines that the magnified parameter value range is 0-60 (because the minimum parameter value is 0, it cannot be reduced to -20).

[0245] Accordingly, the iteration termination conditions in the parameter value optimization phase can include scaling number conditions and error conditions.

[0246] In one possible implementation, when the number of scaling operations in the subrange of parameter values ​​reaches a threshold, or when the minimum photoresist pattern error in the third model set is less than an error threshold, the computer device determines the parameter value of the second sample photoresist model corresponding to the minimum photoresist pattern error in the third model set as the target parameter value.

[0247] In this embodiment, the computer device avoids getting stuck in local optima during the parameter value search process by scaling the parameter value range and sampling the second sample photoresist model within the scaled parameter value range, and avoids the search being too slow due to an excessively large search range, which helps to improve the efficiency and quality of parameter value optimization.

[0248] In an illustrative example, such as Figure 8 As shown, the computer device first determines the initial model set through cluster sampling, that is, based on the clustering results of the model features (parameter values) of the second sample photoresist model 801, dividing the second sample photoresist model 801 into a third model set 802 and a fourth model set 803. After completing the model set initialization, the computer device determines the basic term coefficients based on the second sample photoresist model 801 in the third model set 802 using the constrained least squares method. Then, it uses the second sample photoresist model 801 with the basic term coefficients to predict the photolithographic pattern, obtains the predicted photoresist pattern, and performs actual photolithography using photolithography equipment to obtain the actual photoresist pattern, thereby determining the photoresist pattern error.

[0249] Furthermore, the computer equipment dynamically scales the parameter value range and resamples based on the changes in the minimum photoresist pattern error during the iterative update process, thereby improving the efficiency and accuracy of parameter value optimization.

[0250] When the number of scaling operations in the subrange of parameter values ​​does not reach the threshold and the minimum photoresist pattern error does not reach the error threshold, the computer device trains a Gaussian process regression model 804 based on the model features of the models in the current third model set 802 and the photoresist pattern error, and applies the Gaussian process regression model 804 to the model optimization process.

[0251] During the model optimization process, the computer device applies the Gaussian process regression model 804 to the models in the fourth model set 803 to obtain the mean and uncertainty representing the error distribution 805. Further, based on the mean and uncertainty, the computer device determines the expected improvement value 806 corresponding to the second sample photoresist model 801 in the fourth model set 803, and then adds the second sample photoresist models 801 with the largest expected improvement values ​​806 to the third model set 802.

[0252] When the number of scaling operations in the sub-range of parameter values ​​reaches the threshold, or when the minimum photoresist pattern error reaches the error threshold, the computer device stops iteratively updating the third model set 802 and determines the parameter value corresponding to the minimum photoresist pattern error in the current third model set 802 as the target parameter value.

[0253] To verify the effectiveness of the technical solution provided in the embodiments of this application, the photoresist pattern prediction effects of the target photoresist model determined by the photoresist model determination method provided in the embodiments of this application (hereinafter referred to as the solution of this application) and the photoresist model determined under a fixed mathematical form method (hereinafter referred to as the comparison solution) are compared. The comparison results are as follows: Figure 9 As shown. From Figure 9 It can be seen that the EDE of the photoresist model obtained by using the scheme of this application is within 1nm, and is significantly smaller than the EDE of the photoresist model obtained by using the comparative scheme.

[0254] Furthermore, from Figure 10 It is evident that the photoresist pattern predicted by the photoresist model determined using the scheme of this application is almost completely consistent with the outline of the actual photoresist pattern.

[0255] Figure 11 This is a structural block diagram of a photoresist model determination apparatus provided in an exemplary embodiment of this application. The apparatus includes:

[0256] The acquisition module 1101 is used to acquire model information of the photoresist model. The model information includes the basic items of the photoresist model, the range of the number of basic items of the basic items, and the range of parameter values ​​of the parameters in the basic items. The photoresist model is used to predict the photoresist pattern formed by photolithography.

[0257] The sampling module 1102 is used to sample a sample photoresist model based on the model information, wherein the number of basic items in the basic items of the sample photoresist model is within the range of the number of basic items, and the parameter values ​​of the parameters in the basic items are within the range of the parameter values.

[0258] The determination module 1103 is used to determine the number of target basic items and the target parameter value based on the photoresist pattern error corresponding to the sample photoresist model. The photoresist pattern error is the error between the predicted photoresist pattern obtained by the sample photoresist model and the actual photoresist pattern.

[0259] In some embodiments, the determining module 1103 is configured to:

[0260] Based on the photoresist pattern error corresponding to the first sample photoresist model, the number of target basic items is determined. The number of basic items of the basic items in the first sample photoresist model is within the range of the number of basic items, and the parameter values ​​of the parameters in the basic items are within the range of the parameter values.

[0261] Based on the photoresist pattern error corresponding to the second sample photoresist model, the target parameter value is determined. The number of basic items in the second sample photoresist model is the target number of basic items, and the parameter value of the parameter in the basic item is within the range of the parameter value.

[0262] In some embodiments, during the process of determining the number of target basic items based on the photoresist pattern error corresponding to the first sample photoresist model, the determining module 1103 is used to:

[0263] The first sample photoresist model is divided into a first model set and a second model set;

[0264] Based on the model features of the first sample photoresist model in the first model set, and the photoresist pattern error corresponding to the first sample photoresist model in the first model set, a mapping relationship between model features and errors is constructed. The model features include the number of basic terms and the parameter values.

[0265] Based on the mapping relationship, the first sample photoresist model is selected from the second model set and updated to the first model set;

[0266] The mapping relationship is updated based on the iteratively updated first model set;

[0267] When the number of iterations of the first model set reaches the first iteration threshold, the number of basic terms of the first sample photoresist model corresponding to the minimum photoresist pattern error in the first model set is determined as the target number of basic terms.

[0268] In some embodiments, the determining module 1103 is further configured to:

[0269] Clustering the model features of the first sample photoresist model yields multiple clusters;

[0270] Add the first sample photoresist model corresponding to the model feature closest to the cluster center in each cluster to the first model set;

[0271] Add the first sample photoresist model, which is not in the first model set, to the second model set.

[0272] In some embodiments, the determining module 1103 is configured to:

[0273] Based on the model features of the first sample photoresist model in the first model set, and the photoresist pattern error corresponding to the first sample photoresist model in the first model set, a Gaussian process regression model is constructed. The input of the Gaussian process regression model is the model features, and the output is the distribution of the photoresist pattern error. The distribution of the photoresist pattern error is represented by the mean and uncertainty.

[0274] In some embodiments, the determining module 1103 is configured to:

[0275] Based on the Gaussian process regression model, the mean and uncertainty corresponding to the model features of the first sample photoresist model in the second model set are determined;

[0276] Based on the mean and the uncertainty, the expected improvement value corresponding to the first sample photoresist model in the second model set is determined, and the expected improvement value is used to characterize the quality of the first sample photoresist model.

[0277] Based on the expected improvement value, the first sample photoresist model is selected from the second model set and added to the first model set.

[0278] In some embodiments, the determining module 1103 is further configured to:

[0279] Based on the first sample photoresist model in the first model set, the coefficients of each of the basic terms are determined by the least squares method.

[0280] The predicted photoresist pattern is obtained by using the first sample photoresist model with the aforementioned basic term coefficients.

[0281] The error between the predicted photoresist pattern and the actual photoresist pattern is determined as the photoresist pattern error corresponding to the first sample photoresist model.

[0282] In some embodiments, the sampling module 1102 is configured to:

[0283] Based on the range of the number of the basic items, multiple model feature spaces are determined, and different model feature spaces have different numbers of basic items.

[0284] Based on the parameter value range, the first sample photoresist model is obtained by sampling in each of the model feature spaces. The first sample photoresist model is obtained by random sampling or Latin hypercube sampling.

[0285] In some embodiments, during the process of determining the target parameter value based on the photoresist pattern error corresponding to the second sample photoresist model, the determining module 1103 is used to:

[0286] The second sample photoresist model is divided into a third model set and a fourth model set;

[0287] Based on the model features of the second sample photoresist model in the third model set, and the photoresist pattern error corresponding to the second sample photoresist model in the third model set, a mapping relationship between features and errors is constructed. The model features include the number of basic terms and the parameter values.

[0288] Based on the mapping relationship, the second sample photoresist model is selected from the fourth model set and updated to the third model set;

[0289] The mapping relationship is updated based on the iteratively updated third model set;

[0290] If the iteration termination condition is met, the parameter value of the second sample photoresist model corresponding to the minimum photoresist pattern error in the third model set is determined as the target parameter value.

[0291] In some embodiments, the sampling module 1102 is configured to:

[0292] Based on the number of the target basic terms, the model feature space is determined;

[0293] Based on the parameter value range, the second sample photoresist model is obtained by sampling in the model feature space through range scaling. The second sample photoresist model is obtained by random sampling or Latin hypercube sampling.

[0294] In some embodiments, the sampling module 1102 is configured to:

[0295] Based on the parameter value sub-range, the second sample photoresist model is obtained by sampling in the model feature space, wherein the parameter value sub-range belongs to the parameter value range;

[0296] When the number of iterations in the third model set reaches the second iteration threshold, and the minimum photoresist pattern error in the third model set remains unchanged, the sub-range of parameter values ​​is narrowed, and the narrowed sub-range of parameter values ​​belongs to the parameter value range.

[0297] When the number of iterations in the third model set reaches the second iteration threshold and the minimum photoresist pattern error in the third model set decreases, the sub-range of parameter values ​​is enlarged, and the enlarged sub-range of parameter values ​​belongs to the parameter value range.

[0298] In some embodiments, the determining module 1103 is configured to:

[0299] Centering on the parameter value of the second sample photoresist model corresponding to the smallest photoresist pattern error in the third model set, the sub-range of the parameter value is narrowed;

[0300] Centered on the parameter value of the second sample photoresist model corresponding to the smallest photoresist pattern error in the third model set, the sub-range of the parameter value is expanded.

[0301] In some embodiments, the determining module 1103 is configured to:

[0302] If the number of scaling operations within the sub-range of the parameter value reaches a threshold, or if the minimum photoresist pattern error in the third model set is less than an error threshold, the parameter value of the second sample photoresist model corresponding to the minimum photoresist pattern error in the third model set is determined as the target parameter value.

[0303] In some embodiments, the determining module 1103 is further configured to:

[0304] Clustering the model features of the second sample photoresist model yields multiple clusters;

[0305] Add the second sample photoresist model corresponding to the model feature closest to the cluster center in each cluster to the third model set;

[0306] The second sample photoresist model, which is not in the third model set, is added to the fourth model set.

[0307] In some embodiments, the determining module 1103 is configured to:

[0308] Based on the model features of the second sample photoresist model in the third model set, and the photoresist pattern error corresponding to the second sample photoresist model in the third model set, a Gaussian process regression model is constructed. The input of the Gaussian process regression model is the model features, and the output is the distribution of the photoresist pattern error. The distribution of the photoresist pattern error is represented by the mean and uncertainty.

[0309] In some embodiments, the determining module 1103 is configured to:

[0310] Based on the Gaussian process regression model, the mean and uncertainty corresponding to the model features of the second sample photoresist model in the fourth model set are determined;

[0311] Based on the mean and the uncertainty, the expected improvement value corresponding to the second sample photoresist model in the fourth model set is determined, and the expected improvement value is used to characterize the quality of the second sample photoresist model.

[0312] Based on the expected improvement value, the second sample photoresist model is selected from the fourth model set and added to the third model set.

[0313] In some embodiments, the determining module 1103 is further configured to:

[0314] Based on the second sample photoresist model in the third model set, the coefficients of each basic term are determined by the least squares method.

[0315] The predicted photoresist pattern is obtained by using the second sample photoresist model with the aforementioned basic term coefficients.

[0316] The error between the predicted photoresist pattern and the actual photoresist pattern is determined as the photoresist pattern error corresponding to the second sample photoresist model.

[0317] In some embodiments, the base terms of the photoresist model include at least one of Ag, Mg, Acid, Base, and Grad terms;

[0318] The Ag term includes the standard deviation of the Ag Gaussian distribution;

[0319] The Mg term includes the standard deviation of the Gaussian distribution of Mg;

[0320] The Acid term includes the standard deviation of the Gaussian distribution of Acid and the Acid cutoff threshold.

[0321] The Base term includes the standard deviation of the Base Gaussian distribution and the Base cutoff threshold;

[0322] The Grad term includes the standard deviation of the Grad Gaussian distribution.

[0323] Please refer to Figure 12 The diagram illustrates a structural schematic of a computer device provided in an exemplary embodiment of this application.

[0324] Computer device 1200 includes a central processing unit (CPU) 1201, a system memory 1204 including random access memory 1202 and read-only memory 1203, and a system bus 1205 connecting the system memory 1204 and the CPU 1201. Computer device 1200 also includes a basic input / output system (I / O system) 1206 to facilitate information transfer between various devices within the computer, and a mass storage device 1207 for storing the operating system 1213, application programs 1214, and other program modules 1215.

[0325] The basic input / output system 1206 includes a display 1208 for displaying information and an input device 1209 for user input, such as a mouse or keyboard. Both the display 1208 and the input device 1209 are connected to the central processing unit 1201 via an input / output controller 1210 connected to the system bus 1205. The basic input / output system 1206 may also include the input / output controller 1210 for receiving and processing input from multiple other devices such as a keyboard, mouse, or electronic stylus. Similarly, the input / output controller 1210 also provides output to a display screen, printer, or other types of output devices.

[0326] The mass storage device 1207 is connected to the central processing unit 1201 via a mass storage controller (not shown) connected to the system bus 1205. The mass storage device 1207 and its associated computer-readable media provide non-volatile storage for the computer device 1200. That is, the mass storage device 1207 may include computer-readable media (not shown) such as a hard disk or drive.

[0327] Without loss of generality, the computer-readable medium may include computer storage media and communication media. Computer storage media include volatile and non-volatile, removable and non-removable media implemented using any method or technology for storing information such as computer-readable instructions, data structures, program modules, or other data. Computer storage media include random access memory (RAM), read-only memory (ROM), flash memory or other solid-state storage technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic tape cassettes, magnetic tape, disk storage, or other magnetic storage devices. Of course, those skilled in the art will recognize that the computer storage media are not limited to the above-mentioned types. The system memory 1204 and mass storage device 1207 described above can be collectively referred to as memory.

[0328] The memory stores one or more programs, which are configured to be executed by one or more central processing units 1201. The one or more programs contain instructions for implementing the methods described above, and the central processing unit 1201 executes the one or more programs to implement the methods provided in the various method embodiments described above.

[0329] According to various embodiments of this application, the computer device 1200 can also be connected to a remote computer on a network, such as the Internet. That is, the computer device 1200 can be connected to the network 1212 via the network interface unit 1211 connected to the system bus 1205, or the network interface unit 1211 can be used to connect to other types of networks or remote computer systems (not shown).

[0330] This application also provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the method for determining the photoresist model provided in the above embodiments.

[0331] Optionally, the computer-readable storage medium may include ROM, RAM, solid-state drives (SSDs), or optical discs, etc. The RAM may include resistive random access memory (ReRAM) and dynamic random access memory (DRAM).

[0332] This application provides a computer program product or computer program that includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the photoresist model determination method described in the above embodiments.

[0333] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0334] The above description is merely an optional embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for determining a photoresist model, characterized in that, The method comprises: obtaining model information of a photoresist model, the model information comprising a basis term of the photoresist model, a basis term quantity range of the basis term, and a parameter value range of a parameter in the basis term, the photoresist model being used for predicting a photoresist pattern formed by photolithography; based on the model information, sampling to obtain a sample photoresist model, a basis term quantity of the basis term in the sample photoresist model being located in the basis term quantity range, and a parameter value of the parameter in the basis term being located in the parameter value range; based on a photoresist pattern error corresponding to the sample photoresist model, determining a target basis term quantity and a target parameter value, the photoresist pattern error being an error between a predicted photoresist pattern obtained by the sample photoresist model and an actual photoresist pattern.

2. The method of claim 1, wherein, The method comprises: based on the photoresist pattern error corresponding to the first sample photoresist model, determining the target basis term quantity, the basis term quantity of the basis term in the first sample photoresist model being located in the basis term quantity range, and the parameter value of the parameter in the basis term being located in the parameter value range; based on the photoresist pattern error corresponding to the second sample photoresist model, determining the target parameter value, the basis term quantity of the basis term in the second sample photoresist model being the target basis term quantity, and the parameter value of the parameter in the basis term being located in the parameter value range.

3. The method of claim 2, wherein, The method comprises: dividing the first sample photoresist model into a first model set and a second model set; based on a model feature of the first sample photoresist model in the first model set and the photoresist pattern error corresponding to the first sample photoresist model in the first model set, constructing a mapping relationship between the model feature and the error, the model feature comprising the basis term quantity and the parameter value; based on the mapping relationship, selecting the first sample photoresist model from the second model set and updating to the first model set; based on the first model set after iterative updating, updating the mapping relationship; in a case where an iterative updating number of the first model set reaches a first iteration number threshold, determining the basis term quantity of the first sample photoresist model corresponding to the smallest photoresist pattern error in the first model set as the target basis term quantity.

4. The method of claim 3, wherein, The method comprises: clustering the model features of the first sample photoresist model to obtain a plurality of clusters; adding the first sample photoresist model corresponding to the model feature closest to the cluster center in each cluster to the first model set; adding the first sample photoresist model other than the first model set to the second model set.

5. The method of claim 3, wherein, The model features of the first sample photoresist model in the first model set and the photoresist pattern error corresponding to the first sample photoresist model in the first model set are used to construct a mapping relationship between model features and errors, including: A Gaussian process regression model is constructed based on the model features of the first sample photoresist model in the first model set and the photoresist pattern error corresponding to the first sample photoresist model in the first model set, the input of the Gaussian process regression model is the model feature, the output is the distribution of the photoresist pattern error, and the distribution of the photoresist pattern error is expressed by mean and uncertainty.

6. The method of claim 5, wherein, The first sample photoresist model is selected from the second model set based on the mapping relationship and added to the first model set, including: Based on the Gaussian process regression model, the mean and uncertainty corresponding to the model features of the first sample photoresist model in the second model set are determined; Based on the mean and the uncertainty, the expected improvement value corresponding to the first sample photoresist model in the second model set is determined, which is used to represent the quality of the first sample photoresist model; Based on the expected improvement value, the first sample photoresist model is selected from the second model set and added to the first model set.

7. The method of claim 3, wherein, The method further includes: Based on the first sample photoresist model in the first model set, the basis term coefficients corresponding to each basis term are determined by least squares method; The predicted photoresist pattern is obtained by using the first sample photoresist model with the basis term coefficients. The error between the predicted photoresist pattern and the actual photoresist pattern is determined as the photoresist pattern error corresponding to the first sample photoresist model.

8. The method of claim 3, wherein, Based on the model information, sample photoresist models are obtained, including: Based on the basis term number range of the basis term, a plurality of model feature spaces are determined, and different model feature spaces have different basis term numbers of the basis term; Based on the parameter value range, the first sample photoresist model is sampled in each model feature space, and the first sample photoresist model is obtained by random sampling or Latin hypercube sampling.

9. The method of claim 2, wherein, Based on the photoresist pattern error corresponding to the second sample photoresist model, the target parameter value is determined, including: The second sample photoresist model is divided into a third model set and a fourth model set; Based on the model features of the second sample photoresist model in the third model set and the photoresist pattern error corresponding to the second sample photoresist model in the third model set, a mapping relationship between features and errors is constructed, and the model features include the basis term number and the parameter value; Based on the mapping relationship, the second sample photoresist model is selected from the fourth model set and updated to the third model set; Based on the iteratively updated third model set, the mapping relationship is updated; In a case where the iteration end condition is met, a parameter value of the second sample photoresist model corresponding to a minimum photoresist pattern error in the third model set is determined as the target parameter value.

10. The method of claim 9, wherein, The sampling to obtain the sample photoresist model based on the model information comprises: determining a model feature space based on the target basis term quantity; sampling the second sample photoresist model in the model feature space by range scaling based on the parameter value range, the second sample photoresist model being obtained by random sampling or Latin hypercube sampling.

11. The method of claim 10, wherein, The sampling to obtain the second sample photoresist model in the model feature space by range scaling based on the parameter value range comprises: sampling the second sample photoresist model in the model feature space based on a parameter value sub-range, the parameter value sub-range belonging to the parameter value range; in a case where the iteration update number of the third model set reaches a second iteration update number threshold and the minimum photoresist pattern error in the third model set remains unchanged, reducing the parameter value sub-range, the reduced parameter value sub-range belonging to the parameter value range; in a case where the iteration update number of the third model set reaches the second iteration update number threshold and the minimum photoresist pattern error in the third model set decreases, enlarging the parameter value sub-range, the enlarged parameter value sub-range belonging to the parameter value range.

12. The method of claim 11, wherein, The reducing of the parameter value sub-range comprises: reducing the parameter value sub-range with the parameter value of the second sample photoresist model corresponding to the minimum photoresist pattern error in the third model set as the center; The enlarging of the parameter value sub-range comprises: enlarging the parameter value sub-range with the parameter value of the second sample photoresist model corresponding to the minimum photoresist pattern error in the third model set as the center.

13. The method of claim 11, wherein, The determination of the parameter value of the second sample photoresist model corresponding to the minimum photoresist pattern error in the third model set as the target parameter value in a case where the iteration end condition is met comprises: in a case where the scaling number of the parameter value sub-range reaches a number threshold or the minimum photoresist pattern error in the third model set is less than an error threshold, determining the parameter value of the second sample photoresist model corresponding to the minimum photoresist pattern error in the third model set as the target parameter value.

14. The method of claim 9, wherein, The division of the second sample photoresist model into the third model set and the fourth model set comprises: clustering the model features of the second sample photoresist model to obtain multiple clusters; adding the second sample photoresist model corresponding to the model feature closest to the cluster center in each cluster to the third model set; adding the second sample photoresist model other than the third model set to the fourth model set.

15. The method of claim 9, wherein, The construction of a mapping relationship between features and errors based on the model features of the second sample photoresist model in the third model set and the photoresist pattern errors corresponding to the second sample photoresist model in the third model set comprises: constructing a Gaussian process regression model based on the model features of the second sample photoresist model in the third model set and the photoresist pattern error corresponding to the second sample photoresist model in the third model set, wherein the input of the Gaussian process regression model is the model features, the output is the distribution of the photoresist pattern error, and the distribution of the photoresist pattern error is expressed by a mean value and an uncertainty.

16. The method of claim 15, wherein, The selecting and updating of the second sample photoresist model from the fourth model set to the third model set based on the mapping relationship comprises: determining the mean value and the uncertainty corresponding to the model features of the second sample photoresist model in the fourth model set based on the Gaussian process regression model; determining an expected promotion value corresponding to the second sample photoresist model in the fourth model set based on the mean value and the uncertainty, wherein the expected promotion value is used to represent the quality of the second sample photoresist model; selecting and adding the second sample photoresist model from the fourth model set to the third model set based on the expected promotion value.

17. A device for determining a photoresist model, characterized in that, The device comprises: an acquisition module configured to acquire model information of a photoresist model, wherein the model information comprises a basis term of the photoresist model, a basis term quantity range of the basis term, and a parameter value range of a parameter in the basis term, and the photoresist model is used to predict a photoresist pattern formed by lithography; a sampling module configured to sample to obtain a sample photoresist model based on the model information, wherein the basis term quantity of the basis term in the sample photoresist model is located in the basis term quantity range, and the parameter value of the parameter in the basis term is located in the parameter value range; a determination module configured to determine a target basis term quantity and a target parameter value based on a photoresist pattern error corresponding to the sample photoresist model, wherein the photoresist pattern error is an error between a predicted photoresist pattern predicted by the sample photoresist model and an actual photoresist pattern.

18. A computer device, comprising: The computer device comprises a processor and a memory, and the memory stores at least one computer instruction, which is loaded and executed by the processor to implement the photoresist model determination method according to any one of claims 1 to 16.

19. A computer-readable storage medium, characterized in that, The readable storage medium stores at least one computer instruction, which is loaded and executed by the processor to implement the photoresist model determination method according to any one of claims 1 to 16.

20. A computer program product, characterised in that, The computer program product comprises computer instructions stored in a computer readable storage medium, and the processor acquires the computer instructions from the computer readable storage medium, and executes the computer instructions to implement the photoresist model determination method according to any one of claims 1 to 16.

Images